We study a moral hazard model with a risk-neutral principal and risk-neutral agent, in which the output is stochastically determined by an agent's hidden effort and an uncertain state. We examine how the contractibility of ex-post state observations shapes the principal's optimal informationprovision ex-ante. The principal faces a trade-off: finer information allows for effort levels to be tailored to the realized state (information efficiency), whereas coarser information enables the principal to aggregate the agent's incentives across states to reduce agency costs (cost-saving effect). Undercomplete contracts where the state is ex-post contractible, the cost-saving effect is first order, and it always dominates the second-order information efficiency. Conversely, under incomplete contracts where the state is not ex-post contractible, the cost-saving effect becomes second order, and its magnitude can be smaller than the second-order information efficiency. With binary output, if the production function is additive-separable, then complete information is optimal under incomplete contracts, while no information is optimal under complete contracts. The same insight preserves with a risk-averse agent. (joint with Gregorio Curello and Yimeng Zhang).